To the editor:
I’m sympathetic to the general thrust of Steven Mintz’s argument in Inside Increased Ed, “Writing within the Age of AI Suspicion” (April 2, 2025). AI-detection applications are unreliable. To the diploma that instructors depend on AI detection, they contribute to the erosion of belief between instructors and college students—not an excellent factor. And since AI “detection” works by assessing issues just like the smoothness or “fluency” in writing, they implicitly invert our values: We’re tempted to have larger regard for much less structured or coherent writing, because it strikes us as extra genuine.
Mintz’s article is probably deceptive, nevertheless. He repeatedly testifies that in testing the detection software program, his and different non-AI-produced writing yielded sure scores as “p.c AI generated.” For example, he writes, “27.5 p.c of a January 2019 piece … was deemed more likely to comprise AI-generated textual content.” Though the software program Mintz used for this train (ZeroGPT) does declare to establish “how a lot” of the writing it flags as AI-generated, many different AI detectors (e.g., chatgptzero) point out relatively the diploma of likelihood that the writing as a complete was written by AI. Each varieties of information are imperfect and problematic, however they convey various things.
Once more, Mintz’s argument is beneficial. But when conscientious instructors are going to take a stand towards applied sciences on empirical or principled grounds, they’ll do effectively to exhibit appreciation for the nuances of the assorted instruments.
